Efficiency optimization of induction motor (IM) drives is a major subject based on these drives' extensive use in the industry. Among the different proposed methods, a model-based approach (MBA) seems to be the fast one. However, this method needs the motor parameters that must be correctly identified. On the other hand, a search-based approach (SBA) is a parameter-independent method but needs a greater convergence time. In this paper, a novel model-based loss-minimization approach is presented, which is combined with a backstepping direct torque control of the IM drive. An improved search-based method for efficiency optimization is also introduced. The proposed controller is realized in the stationary reference frame and has a fast-tracking capability of rotor flux and electromagnetic torque. Moreover, a sliding-mode rotor-flux observer is introduced, which is employed for simultaneous determination of rotor-flux space vector, rotor speed, and rotor time constant. The proposed control idea is experimentally implemented in real time using a field-programmable gate-array board synchronized with a personal computer. Simulation and experimental results are finally presented to verify the effectiveness of the method proposed.